Two Notable New Forecasting Texts Principles Of Business .

2y ago
259.65 KB
6 Pages
Last View : 6d ago
Last Download : 5m ago
Upload by : Ellie Forte

Book ReviewsTwo Notable New Forecasting TextsPrinciples of Business Forecasting by Keith Ord & Robert FildesForecasting: Principles and Practiceby Rob Hyndman & George AthanasopoulosReviewed by Stephan KolassaPrinciples of BusinessForecasting (PoBF)Keith Ord of Georgetown Universityand Robert Fildes of LancasterUniversity are well-known figuresin forecasting. Both have publishedextensively and have considerableexperience in educating tomorrow’sforecasters and in hands-on projects in business forecasting. Nowthey have joined forces to write anew textbook: Principles of BusinessForecasting (PoBF; Ord & Fildes, 2013), a506-page tome full of forecasting wisdom.Coverage and SequencingPoBF follows a commonsense order, startingout with chapters on the why, how, and basictools of forecasting. Exponential Smoothingcomes next, the workhorse of business forecasters, as described in the methods tutorialin this issue. The treatment of smoothingis spread over three chapters: one for nonseasonal data, one on seasonal smoothing(other approaches to decomposing seasonaltime series are also touched upon here), andone on the relationship between smoothingand state space models. State space models,which allow prediction intervals to be calculated for exponential smoothing procedures,are a relatively recent development, and theauthors succeed in presenting this normallycomplex topic simply and clearly.PoBF then moves on to regression-type forecasting methods, with a chapter on ARIMA(Box-Jenkins models), a chapter on simplelinear regression, a chapter on multiple linearregression, and another on model building.40 FORESIGHT Fall 2012The model-building chapter provides useful extensions of regression modeling tocover indicator (dummy) variables; laggedvariables and errors; procedures for variable selection (such as stepwise regression);the problems of multicollinearity, changingvariance, and structural change; nonlinearterms and models; and, finally, outliers andleverage points.We next find a chapter on advanced methodsof forecasting (predictive classification, neural networks, and vector autoregression). Forthis complex material, the authors’ presentation is down-to-earth and informative, withhelpful numerical examples of the concepts.The final three chapters give the text a truepractitioner focus. They examine judgmentalforecasting, forecasting in the larger businesscontext, and the processes and politics offorecasting. An extensive discussion of forecasting support systems offers valuable perspectives to the business forecaster. In thesechapters, Ord and Fildes have given us anup-to-date overview of the way forecasting isconducted in the business organization.Level of SophisticationPoBF is not always an easy book – but thenagain, neither is forecasting easy. It containsa fair amount of mathematics and statisticsin explaining the forecasting methods, butadmirably avoids the intricate details forwhich we rely on modern statistical forecasting packages, and provides the interestedreader with pointers to other literature.The book is not meant to be a compendium ofacademic knowledge on forecasting. Rather,

it is aimed at practicing business forecasters, who should have a certain knowledgeabout the underlying statistics in their forecasting package, but who will rarely need touse mathematical formulations to write newprocedures (e.g., to maximize likelihoods byhand).That said, how much math and statistics willa reader need to profit from PoBF? Actually,surprisingly little. Higher mathematics (linear algebra and calculus) are used very sparingly, and one can skip these derivations withlittle loss. Most of the book – specifically,everything apart from the ARIMA, regression, and model-building chapters – is veryaccessible to readers without a backgroundin statistics. For the methodological chapters,some background in statistics is extremelyhelpful, at least to the extent of understanding Q-Q plots, p-values, null hypothesis significance testing, and various distributions(e.g., z, t, F, χ2). When operational, the book’swebsite will contain a primer on basic statistical concepts along with the data sets usedand a glossary.I found PoBF to be extremely readable. Inparticular, I like the way it spreads the moretechnical topics over multiple chapters.The discussion questions, minicases, andexercises are very helpful in understanding each chapter. Each chapter closes witha list of principles that succinctly summarize the takeaways. They credit this idea toScott Armstong’s Principles of Forecastinghandbook (2002), a volume they refer to frequently.SummaryPoBF is perfect for the practicing (as opposedto the academic) forecaster. It draws attentionto the fact that forecasts, targets, and plansare three very different things. It describesbiases within organizations due to varyingincentives of the different functional areas. Itdiscusses surveys about the biggest problemsforecasters face, and offers recommendationson how to address them. It explains how toselect forecasting software and how to marryjudgment with statistical forecasts.Is there anything here I would have done differently? Of course. I found it unclear whycertain topics (unit roots, cointegration, simulations to assess uncertainty) ended up inappendices or nontechnical chapters whenthey would have fit well into the advancedmethods chapter. I would have liked moreattention paid to three topics most forecasters seem to agree are useful to improvingforecasts: using simple methods as benchmarks, combining forecasts, and dampening trends. All three are mentioned, but thecasual reader can be excused for missingthem. Lastly, cost of forecast error and forecast value-added concepts are lacking.These are extremely minor criticisms. Iwholeheartedly recommend Principles ofBusiness Forecasting to every business forecaster, to everyone who aspires to become abusiness forecaster, and to everyone who istraining aspiring business forecasters. (I amhoping that forecast managers will read thisbook as well, but I’m probably being overlyoptimistic.) You will certainly profit fromreading it. I know I did, and I know I will doso again when I reread it.Forecasting: Principles andPractice (FP&P)Rob Hyndman and George Athanasopoulos,both at Monash University, are also wellknown in the forecasting community. Theyhave collaborated to produce an online textbook, Forecasting: Principles and Practice(FP&P), which is explicitly intended asa replacement for Forecasting: Methodsand Applications, the earlier textbook byMakridakis, Wheelwright & Hyndman(1998). FP&P is available free of charge at It is not finished asyet, but the authors hope to finalize it by FORESIGHT41

end of 2012 and to publish a print versionin 2013.OrganizationThe book is at first conventionally organized, beginning with “Getting Started” followed by “The Forecaster’s Toolbox,” whichsupplies the basic graphics, numerical datasummaries, simple forecasting methods, andaccuracy metrics. Next comes a chapter onjudgmental forecasts, which at this point intime is only a placeholder. Most academicforecasters would take up judgmental forecasting after having dealt with statisticalapproaches such as ARIMA and ExponentialSmoothing. However, the first thing a practicing forecaster witnesses in his companyis the frequency of judgmental adjustmentsto statistical forecasts. Thus, I applaud theordering of chapters in this way and hopethat in addition to the discussion of “purely”judgmental approaches (such as Delphi),they will examine judgmental adjustmentsto statistical forecasts as well.The book continues with simple regression,multiple regression, and then time-seriesdecomposition, exponential smoothing, andARIMA. All of these chapters are complete(though exercises have not yet been added),but the next chapters – on advanced methods such as dynamic regression, neural networks, further forecasting methods, demandforecasting, and using forecasting methodsin practice – have yet to be finished.Reliance on RAs with the Ord/Fildes text, FP&P requiresonly a modest mathematical and statisticalbackground. Some sections use matrices or42 FORESIGHT Fall 2012calculus, but these parts are clearly markedand can be skipped. An introductory statistics course would be necessary in orderto derive the full benefit from the ARIMAand regression chapters. Since forecasting isinherently a computational science, we seea lot of equations in some places, especiallythe sections on Exponential Smoothing andARIMA. Throughout FP&P, the authorsuse the statistical environment R, a free,open-source statistical computing environment that can be used for forecasting. Infact, Hyndman and Athanasopoulos wrotean R-package (fpp) as a companion to thise-book, and they provide an appendix witha tutorial, “Using R.” Please note that R is aviable alternative to costly proprietary forecasting software, especially for students orforecasters on a budget, as well as havingappeal to real power users since every function can, in principle, be accessed, examined,or modified. However, R lacks the polishedgraphical user interface of proprietary software and can be intimidating for the casualuser (Kolassa & Hyndman, 2010). Yet thepossibility of copying the R code snippetsfrom the e-book and directly pasting theminto an R console to see the output and toexperiment with R is extremely appealing,and will certainly invite more hands-onpracticing; additionally, one can easily usethe R code snippets as blueprints to apply toone’s own data.Unfortunately, the “Using R” tutorial will beof limited help for R novices; for instance, itinexplicably does not mention the “?” command for getting help, or “?” for free search,or the various ways of getting help online thatare crucial for the new user. One exercise inthe book asks the user to explore some ofthe other time series in certain R packages,without mentioning how one would findout what data sets there are in a package.When asked about this, Rob Hyndman commented that it was a conscious decision inorder to force the student to learn how to usethe index; but I am afraid that raising suchhurdles even before the user can start on the

actual exercise (which, remember, was toexamine some time series, not to learn aboutthe help section in R, or the R-help mailinglist or online R tutorials) will only serve tofrighten away anyone who has not alreadyworked with R before. R already has a steeplearning curve; there really is no need to addto it.GraphicsFP&P heavily emphasizes graphical methods. One section in the “Toolbox” chapter isentirely devoted to different ways of plottingdata in time plots, seasonal plots, seasonalsubseries plots, scatterplots, and scatterplotmatrices – and all of these come with thecorresponding R code. This section is oneof the most helpful parts of the book, andevery forecaster would profit from readingit. Looking at one’s data from various anglesalways aids in understanding the underlyingdynamics. Are there trends? Are seasonalpatterns stable over time? Are there outliers? Or entire patterns of suspicious datapoints? One example involves multiple consecutive weeks of zero airplane passengersflying from Melbourne to Sydney becauseof a labor dispute. Such periods should beaccounted for before using these historicaldata to forecast future passenger levels, andthis pattern clearly shows up in the time plot.The next thing I would like to see includedin the section on graphics would be waysto examine entire data sets of hundreds orthousands of time series simultaneously, andthe e-book format would in fact allow suchan updating.OmissionsFP&P supplies almost no references at all,which is a pity. A lot of material is onlytouched on, and the reader could certainlyuse guidance on how and where to delvemore deeply. Not even the IIF, the IJF, orForesight is mentioned, but this can easily berectified in an e-book.As of this writing (September 2012), the bookis not yet comprehensive and there are pointswhere it obviously could be expanded. Forinstance, the section on numerical data summaries explains the mean and the median,the standard deviation, and the interquartilerange – but it does not mention that the meanand the standard deviation are strongly influenced by extreme values and that the medianand the interquartile range are much morerobust. Similarly, there is a good summaryof what multicollinearity is in the chapter onmultiple regression, but no numerical example or guidance on how to actually detect it,e.g., using Variance Inflation Factors or otherestablished collinearity diagnostics. In addition, some chapters are apparently finishedbut lack exercises; this will hopefully be remedied by the end of 2012.eBook FormatThe e-book format has advantages and disadvantages. For one, I don’t need to lug arounda big hard-copy text – I can just open my laptop or my smartphone. The authors can easily correct errors or add material. I can simply copy and paste R code from FP&P to myR console. References are implemented asdirect hyperlinks to online versions of papersor to a book’s page at You cansearch for a term in FP&P using Google witha “” search.Unfortunately, no downloadable versionof the book is available for those long trainrides with limited data connectivity, but FORESIGHT43

authors promise that an offline version willbe offered together with the print version.And it is hard to make notes in an onlinee-book the way one could do in the marginsof a paper textbook, but an offline e-book ina standard format may allow this.SummaryDoes Forecasting: Principles and Practicedeliver on its stated goal of replacingMakridakis, Wheelwright & Hyndman's1998 book? Upon opening the older book,one immediately notices that much of thematerial and also the data sets have beentransferred to the new book. Some material has been cut, e.g., the chapter on longrange forecasting and (unfortunately) thereferences and the pointers to resourcesand forecasting organizations. On the otherhand, much material is updated based onwhat forecasters have learned over the last15 years. Examples include the sections onforecast-accuracy measurement, neural networks, and everything on R. Thus, FP&Preally is an (almost) perfect successor to theolder book.The bottom line: this is the book for youif you are happy with short, to-the-pointexplanations, if you are able to find yourown way through the literature to clarifyunclear points, and if you are comfortableusing R. Everyone else would certainly alsoprofit from leafing through FP&P – the priceis hard to beat, after all – but should preparefor a strenuous read. I have already bookmarked Forecasting: Principles and Practiceand will definitely return to it.COMPARISONWe now have two new books on the market,which both aim to attract the forecastingstudent and the practitioner. How do thesebooks compare?The first note of comparison is the pricetag. Principles of Business Forecasting retailsfor about 150 U.S. on Amazon, whereasForecasting: Principles and Practice is free inits online version. We don’t know yet howmuch the paper or offline versions of FP&Pwill cost.PoBF offers more detailed explanations andillustrations of the methods, and, while longer, examines the methods from differentangles. Overall, it seems to be more gentle inits didactics.FP&P relies exclusively on the free softwarepackage R. PoBF unfortunately does notmention this free computing environment,only referring to commercial statistical orforecasting packages such as Forecast Pro,Minitab, SAS, SPSS, EViews, and PcGive/OxMetrics (except for a brief mention ofthe free specialized econometric packageTRAMO-SEATS).Even apart from the price differential, it ismuch easier for the reader to load the “fpp”package and copy R code from the book tothe R console than it is to go to the PoBFwebsite, download a data set, fire up his orher favorite forecasting program, and loadthe dataset. Thanks to R, the hurdle for actually earning some hands-on experience withforecasting is a lot lower with FP&P than forPoBF.PoBF gives references for further readingafter each chapter, which allows one to learnmore about specific topics. FP&P gives veryfew references, and readers who would like44 FORESIGHT Fall 2012

to learn more about, say, invertibility of anARIMA process would need to hunt for relevant literature themselves.In comparing an e-book and a paper textbook, differences are inherent in the twomedia. FP&P can and will change, and eventhose sections that are already finished canbe rewritten or expanded. Errors and typoscan be corrected (whereas PoBF will publisha list of errata at its website), new researchcan be worked into the text, and readers cansuggest specific topics to be addressed. Thereare no page limits as there were for PoBF. Onthe other hand, as long as no offline versionis available for download, the user cannotannotate his FP&P copy and will need anInternet connection to read it. In contrast, Iwas perfectly able to read my copy of PoBFduring my vacation in an idyllic – and blissfully Internet-free – seaside cottage.One issue on which the two texts do not differ is the level of math/stat knowledge theydemand. Most of both books is accessiblewith no particular math/stat background atall; parts of both books are more easily readif the reader knows matrices and calculus;and readers will need a bit of inferentialstatistics knowledge to fully profit from thechapters on regression and ARIMA.A book for the practicing forecaster alsoneeds to cover the nonscientific “softer”aspects of forecasting: what kind of largerprocesses make use of forecasts, what are theadvantages and disadvantages of locatingthe forecasting function in various organizational units, how about judgmental adjustments to statistical forecasts, and what kindsof incentives do people around the forecaster face? PoBF has very helpful chapters on allthis, while FP&P as yet only has placeholderchapters on these issues.Each book has its strengths, and I would keepboth close to hand; every forecaster shouldat least look at both of them. You may wantto start out with Forecasting: Principles andPractice if you are a cash-strapped student ora forecaster with no budget for a book anda software package, if you are comfortablewith rather terse exposition, if you are happyusing R, if you like to play around with yourdata and your forecasts while you read, and ifyou have a solid Internet connection up andrunning. Conversely, Principles of BusinessForecasting may be the better choice if youhave enough budget (or a library copy of thebook and a campus license for a forecasting package), if you prefer a gentler exposition, if you like to just read your book and/or have the discipline to work with data andsoftware (even though this is slightly moreof a hassle), or if your favorite reading/studying haunt has no Internet connection.But whichever book you start with, it makesa lot of sense to read the other book, too.ReferencesArmstrong, J. Scott, ed. (2002). Principles of Forecasting: A Handbook for Researchers and Practitioners. New York, NY: SpringerHyndman, Robert J. and Athanasopoulos,George (2012). Forecasting: Principles andPractice. Accessed09/04/2012 to 09/07/2012Kolassa, Stephan and Hyndman, Robert J.(2010). Free Open-Source Forecasting Using R.Foresight, 17, 19-24Makridakis, Spyros; Wheelwright, Steve andHyndman, Robert J. ( 1998). Forecasting: Methods and Applications. 3rd edition. Hoboken, NJ:WileyOrd, Keith and Fildes, Robert (2013). Principlesof Business Forecasting. Mason, OH: CengageLearning.Stephan Kolassa is a Senior Re-search Expert at the SAP Center ofExcellence Forecasting and Replenishment and an Associate Editor FORESIGHT45

ects in business forecasting. Now they have joined forces to write a new textbook: Principles of Business Forecasting (PoBF; Ord & Fildes, 2013), a 506-page tome full of forecasting wisdom. Coverage and Sequencing PoBF follows a commonsense order, starting out with chapters on the why, how, and basic tools of forecasting.

Related Documents:

Introduction to Forecasting 1.1 Introduction What would happen if we could know more about the future? Forecasting is very important for: Business. Forecasting sales, prices, inventories, new entries. Finance. Forecasting financial risk, volatility forecasts. Stock prices? Economics. Unemplo

Forecasting with R Nikolaos Kourentzesa,c, Fotios Petropoulosb,c aLancaster Centre for Forecasting, LUMS, Lancaster University, UK bCardi Business School, Cardi University, UK cForecasting Society, This document is supplementary material for the \Forecasting with R" workshop delivered at the International Symposium on Forecasting 2016 (ISF2016).

Importance of Forecasting Make informed business decisions Develop data-driven strategies Create proactive, not reactive, decision making 5 6. 4/28/2021 4 HR & Forecasting “Putting Forecasting in Focus” –SHRM article by Carolyn Hirschman Forecasting Strategic W

Although forecasting is a key business function, many organizations do not have a dedicated forecasting staff, or they may only have a small team. Therefore, a large degree of automation may be required to complete the forecasting process in the time available during each forecasting and planning cycle.

Undoubtedly, this research will enrich greatly the study on forecasting techniques for apparel sales and it is helpful to identify and select benchmark forecasting techniques for different data patterns. 2. Methodology for forecasting performance comparison This research will investigate the performances of different types of forecasting techniques

154 Chapter 2: Informational Texts When you compare and contrast across texts, you look at the similarities and differences in the texts . Comparisons focus on the things that the texts share . Contrasts focus on differences . Comparing and contrasting across texts will help you better understand each text .

Option #1: Operations Management Forecasting Paper . This Critical Thinking assignment option consists of two activities: 1) Performing the Pearson MyOMLab Forecasting Simulation, and 2) a written operations management forecasting paper. Your written operations management forecasting paper

Nutrition of ruminants Developing production systems for ruminants using tropical feed resources requires an understanding of the relative roles and nutrient needs of the two-compartment system represented by the symbiotic relationship between rumen micro-organisms and the host animal. Fibre-rich, low-protein forages and crop residues are the most abundant and appropriate feeds for ruminants .